Published October 19, 2017
Review & Analysis–Soil Physics & Hydrology
Numerical Solution of Richards’ Equation:
A Review of Advances and Challenges
Matthew W. Farthing
Coastal and Hydraulics Lab.
US Army Engineer Research and
Development Center
3909 Halls Ferry Rd
Vicksburg, MS 39180
Fred L. Ogden*
Dep. of Civil and
Architectural Engineering
Univ. of Wyoming
1000 E. Univ. Ave. Dep. 3295
Laramie, WY 82071
Core Ideas
• The numerical solution of Richards’
equation remains challenging.
• Space/time discretization affects both
computational effort and accuracy.
• Adaption of space and time
discretizations produces benefits.
• Dissemination of codes and improved
documentation are needed.
• Recent reformulation of one-
dimensional Richards’ equation
shows promise.
Soil Science Society of America Journal
The flow of water in partially saturated porous media is of importance in fields
such as hydrology, agriculture, environment and waste management. It is also
one of the most complex flows in nature. The Richards’ equation describes the
flow of water in an unsaturated porous medium due to the actions of grav-
ity and capillarity neglecting the flow of the non-wetting phase, usually air.
Analytical solutions of Richards’ equation exist only for simplified cases, so
most practical situations require a numerical solution in one- two- or three-
dimensions, depending on the problem and complexity of the flow situation.
Despite the fact that the first reasonably complete conservative numerical solu-
tion method was published in the early 1990s, the numerical solution of the
Richards’ equation remains computationally expensive and in certain circum-
stances, unreliable. A universally robust and accurate solution methodology has
not yet been identified that is applicable across the range of soils, initial and
boundary conditions found in practice. Existing solution codes have been modi-
fied over years to attempt to increase robustness. Despite theoretical results on
the existence of solutions given sufficiently regular data and constitutive rela-
tions, our numerical methods often fail to demonstrate reliable convergence
behavior in practice, especially for higher-order methods. Because of robust-
ness, the lack of higher-order accuracy and computational expense, alternative
solution approaches or methods are needed. There is also a need for better
documentation of improved solution methodologies and benchmark test prob-
lems to facilitate consistent advances and avoid re-inventing of the wheel.
T
his review paper is intended to serve as a touchstone on the state of the
science of calculating the flow of water through unsaturated porous me-
dia. As active researchers we believe it is good to occasionally take stock
in what has been accomplished up to date, where things may be headed, and what
important issues remain. This review concentrates on computational modeling of
flow through the unsaturated zone via Richards’ equation.
This effort can help provide some focus to the research community and serve
to help propel new research forward, particularly by those just joining the field.
There are many practical problems that require calculation of one-, two-, and three-
dimensional fluxes in the unsaturated zone. For a much broader review of modeling
soil processes, we recommend the recent paper from Vereecken et al. (2016). A
detailed review of mathematical models of infiltration can be found (Assouline,
2013), while Paniconi and Putti (2015) provide historical perspective and an
overview of modeling Richards’ equation in the context of catchment hydrology.
The equation attributed to Richards (1931) that describes the flow of water
through unsaturated porous media under the action of capillarity and gravity was
first published by the English mathematician and physicist Lewis Fry Richardson
in 1922 (Richardson, 1922). Therefore, the equation would rightly be called
“Richardsons’ equation”, although in this paper we continue to call it by its
common name, Richards’ equation.
Soil Sci. Soc. Am. J. 81:1257–1269
doi:10.2136/sssaj2017.02.0058
Received 16 Feb. 2017.
Accepted 4 June 2017.
*Corresponding author: (fogden@uwyo.edu)
© Soil Science Society of America. This is an open access article distributed under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
 Richards’ equation can be seen as a simplification of the
standard two-phase flow formulation for a gas and water phase in
a porous medium where the pressure gradient required to drive
flow of the gas phase is ignored due to the large mobility contrast
between the water and gas phases. One of the most interesting
aspects of the Richards’ equation is that despite its ease of
derivation, it is arguably one of the most difficult equations to
reliably and accurately solve in all of hydrosciences.
Richards’ equation can be written in a number of forms,
namely the water content, mixed water content and capillary
head form, and the head form. In one-dimension, the so-called
“mixed water content form”, because it mixes the water content θ
with the capillary head y(θ), is:
∂q ∂   ∂ y(q)  
=  K (q )  −1    [1]
∂t ∂z   ∂z 
where z is vertical coordinate (positive downward)[L]; t is time
[T]; q equals q(z,t) equals volumetric soil moisture content [-];
y(q) is empirical soil hydraulic capillary head function [L]; K(q)
= empirical unsaturated hydraulic conductivity function [L T–1].
Equation [1] is often written solely as a function of the
water content by introducing the chain rule:
∂q ∂  ∂q 
=  D(q) − K (q)  [2]
∂t ∂z  ∂z  solve using traditional analytical techniques, and it is impossible
where D(θ) = K(θ)(¶y(θ)/¶θ) is referred to as the soil-water
diffusivity [L2 T–1]. As in Eq. [1], the first term in parentheses
captures the effects of capillarity, while second term in
parentheses represents the effect of gravity-driven flux.
In uniform soils the water content or mixed water content
forms of Richards’ equation are valuable because water content
is a continuous variable. However, in nature soils are seldom
uniform over significant length scales, and layered soils are
ubiquitous. In layered soils the water content is discontinuous
across layer interfaces because of unique unsaturated capillary
head relations in the different soil layers (Assouline, 2013).
Rather, the capillary head (y) is continuous, and it is better to
write the Richards’ equation with capillary head as the dependent
variable and evaluate the moisture content in terms of y, θ =
θ(y). This can be done in either a mixed form:
∂ q(y) ∂ 
 ∂y  
−  K ( y)  −1   =
0 [3]
∂t ∂z   ∂z  
or via the chain rule as the fully head-based form:
∂y ∂   ∂ y   [4]
c ( y) − K ( y)  −1   =
0
∂ t ∂ z   ∂z  
where c(y) equals ¶q/¶y equals the specific moisture
capacity [L–1]; K(y) equals the hydraulic conductivity
function written as a function of y[L T–1].
Equations [3] and [4] can both be used to solve unsaturated
and saturated flow problems. One may also encounter Eq. [4] in
a form that includes fluid compressibility effects
1258 Soil Science Society of America Journal
∂y ∂   ∂y  
 S s S a ( y ) + c ( y )  ∂ t − ∂ z  K ( y )  ∂ z −1   =
0 [5]
  
where Ss [L–1] is the specific storage and Sa [-] is the saturation of
the aqueous phase (Miller et al., 1998). While ¶q/¶t = c(y)¶y/¶t
holds at the continuous level through the chain rule, the
nonlinearity of c(y) makes it impossible to maintain this equality
uniformly at the discrete level, and the pure head-based form in Eq.
[4] or [5] is not generally mass conservative. There are techniques
like a chord-slope approximation (Rathfelder and Abriola, 1994)
that can be used in specific cases or flux updating that can promote,
but not ensure, mass conservation (Kirkland et al., 1992). In
general though, high accuracy in the time integration of Eq. [4] is
essential to avoid mass balance errors (Tocci et al., 1997), and Eq.
[3] together with expansion of the nonlinearity θ(y) in the storage
change term, ¶θ/¶t, (Allen and Murphy, 1985; Celia et al., 1990)
is preferred when low-order time integration is used.
Richards’ equation is a nonlinear, degenerate elliptic-
parabolic partial differential equation (List and Radu, 2016).
The head-based forms of the equation, Eq. [3–5], transition
from parabolic to elliptic as the solution domain nears saturation
unless storage effects are included (Ss > 0), in which case Eq. [5]
remains parabolic. The nonlinearity and transition of type in
different portions of the problem domain make it very difficult to
to solve in closed form except for a small number of special cases
(Miller et al., 1998). This difficulty is only exacerbated by the
range of initial and boundary conditions encountered in practice
as well as by the nature of common soil water constitutive relations
that can lead to solutions with low regularity (Alt and Luckhaus,
1983). Specifically, the solution depends on two empirical,
highly nonlinear soil water constitutive relations: the unsaturated
hydraulic conductivity function K(θ) that can be constant or
very near zero for non-positive values of the capillary head, and
the capillary head function y(θ) that can take on arbitrary small
values for relative saturations near 100%. These extremes lead
to degeneracy in the solution of Richards’ equation. The soil
constitutive relations may not be smoothly differentiable at these
extremes and can further have very high slopes and hysteresis,
and be discontinuous at low relative saturations. Furthermore,
infiltration into dry soils often leads to sharp wetting fronts
that result in extremely large spatial gradients of soil hydraulic
properties (Zha et al., 2017). These nonlinearities and the
degeneracy make the design and analysis of numerical schemes for
the Richards’ equation very difficult (Miller et al., 2013).
Exact one-dimensional solutions of Richards’ equation have
been derived for a few specialized forms of the constitutive relations
describing the soil water retention and the unsaturated hydraulic
conductivity functions (Rogers et al., 1983; Broadbridge and
White, 1988; Sander et al., 1988; Barry and Sander, 1991; Barry et
al., 1993; Ross and Parlange, 1994). However, these solutions are
not generally applicable because either the functional forms are
dissimilar from widely used constitutive relations that represent
real soils, and/or the solutions impose strict requirements on
the initial and boundary conditions. These exact solutions are
valuable for verifying numerical solutions.
Given the lack of a general closed-form solution of the
Richards’ equation, numerical solvers are required when accurate
unsaturated flow simulation capability is desired. Of course,
many approximate, empirical, or completely arbitrary methods
have been developed over the years to simulate the flow of
water through unsaturated soils because of the real or perceived
difficulties posed by the numerical solution of Richards’ equation.
Here we ignore those, and we solely consider problems where fully
dynamical simulations in unsaturated porous media are required.
Note that in the following we focus on head-based forms of
Richards’ equation that are valid for unsaturated or fully saturated
(i.e., variably saturated) conditions. When necessary, we will use
the term “strictly unsaturated” to refer to conditions or techniques
that are restricted to relative saturation values less than 100%.
The complexities of solving the Richards’ equation were
discussed by Gray and Hassanizadeh (1991), who referred to the
formulation as “paradoxical and overly simplistic”. Since then,
many authors have presented critiques of Richards’ equation
and existing solution techniques. A far-from-exhaustive sample
includes (Forsyth et al., 1995; Tocci et al., 1997; van Dam and
Feddes, 2000; Vogel et al., 2001; Kavetski et al., 2001a; Farthing
et al., 2003b; Bause and Knabner, 2004; Manzini and Ferraris,
2004; D’Haese et al., 2007; Kees et al., 2008; Vogel and Ippisch,
2008; Juncu et al., 2011; Berninger et al., 2014; Lipnikov et al.,
2016). Our intention here is not to repeat what those authors
have written, rather it is to assess the state of the art, limitations,
needs, and alternatives.
Calculation of Vadose Zone Fluxes
In the Richards’ equation, the physical properties of
the medium are represented as nonlinear coefficients. The
nonlinearities introduce a number of challenges as alluded
to in the introduction. For example, with the widely used van
Genuchten and Mualem constitutive relations, the capillary
head function, y (θ), and the specific moisture capacity, ¶θ/¶y,
approach zero as the moisture content in the porous medium
nears saturation, while the soil-water diffusivity, D(θ), can grow
arbitrarily large (van Genuchten, 1980). Loosely, this behavior
leads to a condition known as degeneracy, when the solution of
Richards’ equation is hampered by coefficients that take values
near zero or infinity. Classical analytical results on the existence,
uniqueness, and regularity of weak solutions can be found
in (Van Duyn and Peletier, 1982; Alt and Luckhaus, 1983),
while Gilding (1991) provides analysis of infiltration in strictly
unsaturated domains, including the positivity and boundedness
of soil moisture and conditions for finite speed of propagation
of wetting fronts. In practice, the difficulties driven by these
nonlinearities often manifest as rapid changes in capillary head
and soil moisture around infiltration fronts as well as complex
material coefficients that are potentially non-continuously
differentiable near full saturation (Miller et al., 1998).
www.soils.org/publications/sssaj 1259
Infiltration is often assumed to be a one-dimensional process
in the vertical direction. Or et al. (2015) demonstrated that higher
dimensional approaches are generally necessary only at scales
less than approximately 10 m. The one-dimensional solution
methodology developed by Celia et al. (1990), which uses modified
Picard iterations to improve mass conservation, has become the
standard numerical approach. It remains essentially the method that
is used in many production codes including the USDA Hydrus-1D
Richards’ equation solver (Simunek et al., 2005).
As a result of the one-dimensional vertical assumption, large-
scale models are able to replace a fully coupled, three-dimensional
solution with many, independent one-dimensional calculations.
Because of reliability issues however, current large-scale models of
land/atmosphere interaction often use approximate or conceptual
methods to model vadose zone fluxes. Those approximations
usually require assumption of runoff generation mechanism,
and are used in land-surface schemes such as: VIC (Liang et al.,
1994), JULES (Cox et al., 1999), NOAH (Ek et al., 2003), CLM
(Niu et al., 2007), NOAH-MP (Yang et al., 2011). The reality
is that the numerical solution of Richards’ equation is avoided
in these schemes because of concerns regarding computational
expense and solution reliability, as well as questions regarding
the appropriateness of applying the equation to represent the
behavior of large model elements or grids.
The penalty in using a Richards’ solver in hydrological
simulation with thousands or millions of one-dimensional solutions
comes from the fact that while many of these may converge with
rapidity and ease, there are conditions where some of them will
not (Niswonger and Prudic, 2009) for a variety of reasons. In those
cases, where convergence is slow or not possible, the code must be
able to identify this condition and move to the next time step with
some approximate solution. This problem affects the reliability of
hydrological simulators that use Richards’ equation. The lack of
solution reliability imposes a large penalty on the use of Richards’
equation solvers when convergence is unattainable. This penalty
takes the form of computational expense, slowdown of parallel
algorithms waiting for detection and approximations in the case
of non-convergent solutions, and solutions with large errors,
including significant mass-balance errors.
The convergence abilities and computational efficiency (or
lack thereof ) for a particular Richards’ equation solver depend
on many factors. The quality of the software engineering and
implementation of the numerics are obviously critical. The
essential numerical components include the spatial and temporal
approximations used to transition from an initial boundary value
problem posed at the continuous level to a discrete system, as well
as the nonlinear and linear algebra techniques that are used to
solve the resulting discrete systems and produce an approximate
solution (Trangenstein, 2013). Here, we focus on the numerical
discretization techniques themselves and not the very important
software engineering issues surrounding the development of high-
quality scientific codes (McConnell, 2004; Bartlett, 2009; Heroux
and Willenbring, 2009). We summarize briefly the current state of
the art and identify outstanding issues for the primary numerical an example, Fig. 2 illustrates a typical piecewise linear Lagrange
components of a Richards’ equation solver below. polynomial, wI, that is the building block for most finite element
models for Richards’ equation. It takes a value of 1 at vertex,
Spatial Discretization I, and is zero at the other vertices of the triangulation, Mh. Of
The overwhelming majority of Richards’ equation solvers course, Eq. [6] and [7] are only intended to provide some context
employ either a finite difference, finite volume, or finite element for the following discussion, and we have neglected any details of
approximation in space. Hybrid combinations are also possible the solution approximation or evaluation of the integrals in Eq.
(Helmig, 2011). Table 1 provides a representative list of research [6] and [7] (Kees et al., 2008).
and production codes along with their spatial approximations. The spatial accuracy of all of these approximation methods
The details of these approaches vary. They all build on a discrete depends on several factors. In particular, it is generally a function
partition or tessellation of the spatial domain of interest. The of the local mesh geometry, which can be expressed through a
tessellation may be a structured grid or more generally an characteristic length scale and a measure of the aspect ratio or
unstructured mesh. Figure 1 illustrates a simple unstructured distortion (Trangenstein, 2013). Loosely, the formal accuracy of
mesh of a polygonal spatial domain, W, consisting of a union of a convergent scheme can be related to the characteristic length
non-empty, non-overlapping triangles, Mh = {We}. scale of the mesh raised to some power, which we will refer to as
Each scheme must then approximate the spatial variation the scheme’s order. There is a traditional rule of thumb that large
of material properties and solution variables on this mesh and angles (hence aspect ratios) negatively impact approximation
enforce the conservation statement and solution dynamics accuracy (Shewchuk, 2002). On the other hand, it is also possible
expressed by Richards’ equation in some way. For example, through mesh optimization techniques to exploit solution
conservation might be enforced point-wise in a traditional finite behavior and gain higher accuracy through the appropriate use of
difference scheme. Or, it might be enforced in an integral sense element anisotropy (Pain et al., 2001; Mostaghimi et al., 2015).
using test functions that are constant over a control volume, V, in The spatial discretization size is relevant beyond just the
a finite volume method formal accuracy. Convergence of a method relates to how well
the scheme approximates a solution to Richards’ equation. It
∂q
∫ d x − ∫ ∇⋅[ K (y)(∇ y −1z ) ] d x = 0 [6] does not address whether or not Richards’ equation is a valid
V ∂t V
approximation for unsaturated flow at the length scales used
Here, 1z is a vector denoting the vertical axis. V could correspond in the mesh, however. For example, it is unlikely that a three-
to a mesh element, We, in “cell-centered” scheme or a dual control dimensional solution of Richards’ equation on a large, multiple-
volume, W  , in a “vertex-centered” scheme (Fig. 1) (Huber and kilometer-scale voxel can accurately represent nature or be
I
Helmig, 2000; Manzini and Ferraris, 2004). Finite element physically meaningful because it will violate the representative
methods are also weighted residual methods that enforce elementary volume assumption (Or et al., 2015).
Richards’ equation weakly, For reference, Fig. 3 shows three commonly used one-
∂q dimensional discretizations. From left to right these spatial
∫ w d x − ∫ ∇⋅[ K (y)(∇ y −1z ) ] w d =
x 0, ∀ w∈W (W) [7] discretizations represent increasing degrees of complexity, and
W ∂t W
increasing degrees of accuracy. The uniform discretization is
Finite element methods impose a richer structure on the space simplest, but the nature of the solution requires fine resolution
of weighting functions, W(W) (Ern and Guermond, 2004). As everywhere. The variable discretization provides high resolution
near the land surface, where accurately simulating the change
Table 1. Alphabetical list of representative research and pro- in water content with time is most important for accurate
duction Richards’ equation codes. partitioning of rainfall or energy (Downer and Ogden, 2004).
Spatial The variable spatial discretization also allows coarsening
Code approximation Reference away from the land surface, which can reduce computational
ADH CG FEM† Howington et al. 1999
CATHY CG FEM Camporese et al., 2010
FEHM CV FEM Zyvoloski, 2007
FEFLOW CG FEM Trefry and Muffels, 2007
HydroGeoSphere CV FEM Brunner and Simmons, 2012
HYDRUS CG FEM Simunek et al., 2008
ParFlow CCD Kollet and Maxwell, 2006
RichardsFOAM FV Orgogozo et al., 2014
TOUGH IFD Finsterle et al., 2008
VS2D CCD Healy, 2008
WASH123D CG FEM Yeh et al., 2011
†C G FEM, Continuous Galerkin Finite Element Method; CV FEM,
Control Volume Finite Element Method; CCD, Cell-Centered
Differences; FV, Finite Volume; IFD, Integrated Finite Differences. Fig. 1. Example triangulation with dual mesh.

1260 Soil Science Society of America Journal

burden without significantly affecting the solution accuracy.
The adaptive spatial discretization is a much more complicated
situation wherein the solver places computational points that
allow the most accurate simulation of the time evolution of the
wetting front.
The past three decades have seen the introduction and
maturation of a number of spatial approximation methods for
elliptic problems that are robust for high contrasts in material
properties and heterogeneity while providing locally conservative
velocities for transport. These schemes include cell and vertex-
centered finite volume schemes, mimetic finite differences,
multi-point flux approximation (MPFA) techniques, mixed
finite elements, and discontinuous Galerkin methods (Chavent
and Roberts, 1991; Arbogast et al., 1995; Aavatsmark et al.,
1998; Campbell et al., 2002; Dawson et al., 2004; Klausen and
Russell, 2004; Trangenstein, 2013).
These techniques have been largely successful for multiphase
flow problems when applied to the global flow (or pressure)
equation in fractional flow formulations (Chavent et al., 1984;
Gerritsen and Durlofsky, 2005; Hoteit and Firoozabadi, 2005;
Chen et al., 2006). They have also been applied to Richards’
equation by a number of researchers (Bergamaschi et al., 1999; Li
et al., 2007a; Arrarás et al., 2009; Klausen et al., 2008; Kumar et
al., 2009). However, their direct translation to variably saturated
flow has proven challenging.
To achieve comparable robustness, non-trivial modifications
like low-order quadrature and mass lumping (Forsyth and
Kropinski, 1997; Arbogast et al., 1998; Woodward and Dawson,
2000; Farthing et al., 2003b; Pop et al., 2004; Belfort et al.,
2009; Younes et al., 2013) have often been required along with
significant work on methods for solving the resulting nonlinear
discrete problems (Paniconi et al., 1991; Bergamaschi et al., 1999;
Bause and Knabner, 2004; List and Radu, 2016). As a result,
cell-centered finite differences and low-order finite volumes/
integrated finite differences, and piecewise linear, continuous
Galerkin finite element schemes remain the dominant schemes
used in practice (Selker and John, 2004; Kollet and Maxwell, 2006;
Yeh et al., 2011; Diersch, 2013). We note that post-processing is
increasingly being used in conjunction with continuous Galerkin
methods to provide velocity fields that are locally conservative
over elemental control volumes (Larson and Niklasson, 2004; Sun
and Wheeler, 2006; Kees et al., 2008; Scudeler et al., 2016).
Local features like steep infiltration fronts make Richards’
equation a natural candidate for adaptive spatial discretizations
Fig. 2. Example piecewise linear weighting function.
www.soils.org/publications/sssaj 1261
like the one sketched on Fig. 3C. Many researchers have shown
that adaptive resolution can greatly improve accuracy and
robustness (Bause and Knabner, 2004; Li et al., 2007b). Indeed
even one-dimensional approximations have shown that they can
pose a significant computational burden and benefit from adaptive
resolution, depending on the combination of soil properties, initial
conditions, and boundary forcing (Miller et al., 2006). However,
despite the introduction of sophisticated adaption techniques
that have have matured in other computational mechanics fields
(Schwab, 1998; Rannacher, 2001; Fidkowski and Darmofal,
2011), including so-called h adaption which modifies the local
mesh spacing (Pettway et al., 2010) as well as methods that
vary the local mesh spacing and approximation order (i.e., h-p
adaption) (Solin and Kuraz, 2011), spatial adaption for Richards’
equation is still not widely used in practice.
Temporal Discretization
As with spatial approximations, solution of time-varying
problems requires partition of the temporal domain of interest.
Generating a discrete partition for an interval, [0,T] 0 = t0 < t1
< ... < tm = T, is much simpler than a complex three-dimensional
domain. Nevertheless, accuracy remains a function of the time
step size. Schemes can be characterized in terms of their local
discretization error, which is the error that would originate over
one time step, Δt(n+1) = t(n+1) –tn, starting from the true solution
at tn. The error that accumulates over the entire interval [0,T] is
known as the global error (Hairer et al., 1996). We will, somewhat
loosely, refer to a method as order p if its local error is order p+1
∈n+1 ≤C (D t n+1 ) p+1 [8]
Here C is a constant that is independent of the time step (Hairer
et al., 1996).
The norm for the numerical solutions of Richards’ equation
is low-order time discretizations, which are globally first-order.
Notable exceptions include the approaches in (Tocci et al.,
1997; Farthing et al., 2003b) based on higher order Backward
Difference Formulas (BDFs) and the second-order Taylor-
Gladwell scheme introduced in (Kavetski et al., 2001a, 2004).
The vast majority of temporal approximations are also implicit
in head-based models that allow fully saturated conditions
to develop in the domain (Celia et al., 1990). Fully implicit
Fig. 3. Three widely used spatial discretizations shown for example
in 1-dimension.
discretizations incorporate the current solution at a time level
into their approximations and so require solution of at least one
nonlinear problem at each time step.
To make the following discussion more precise, we assume a
method of lines (MOL) approach to eliminate partial derivative
terms after applying one of the spatial discretizations from
Section 3. The subsequent discussion would be analogous if
we followed a Rothe paradigm and introduced a temporal
discretization before the spatial discretization (Lang, 2013).
Applying the MOL leads to following semi-discrete problem
when stepping forward from one time level, tn, to the next, tn+1
, y , y ′) 0, for t∈(t n , t n+1 ) [9]
F(t=
Here, yʹÎ Rn is an n-dimensional vector of discrete unknowns
or degrees of freedom resulting from the spatial approximation,
y¢ is the time derivative for y, and the time step is Δt(n+1). Initial
conditions for Eq. [9] are taken from time level n, and we assume
that they are consistent
F(t n , y n , y ′, n ) = 0 [10]
Equation [9] is known as a system of Differential Algebraic
Equations (DAEs) and can be thought of as a generalization of
an Ordinary Differential Equation (ODE) system (Brenan et al.,
1996). As an example, BDFs are multistep methods that can be
thought of as introducing an approximation for y¢,
y ′, n+1 ≈a n+1 y n+1 + b n [11]
and so lead to a discrete nonlinear problem
F ( t n+1 , y=
n+1
, y ′, n+1 ) G=
n+1
( y n+1 ) 0 [12]
that must be solved to obtain y(n+1). Note that Backward Euler is
just the first order BDF
1 1
y ' , n+1 ≈ n+1
y n+1 − n+1 y n [13]
Dt Dt
Similarly, an implicit k stage Runge-Kutta approximation for Eq.
[9] requires solution of the nonlinear problems (Brenan et al.,
1996; Hairer et al., 1996)
( k
)
F t in+1 , y n +D t n+1 ∑ j=1 aij Y ' j , Y ' i =
0, for i =
1,..., k [14]
for the intermediate stage derivatives Y¢i to obtain
y n+=1 y n + ∑ j=1 b j Y ' j [15]
k
Here the implicitness results from one or more coefficient aij >
0, j ≥ i.
The nonlinear system of equations in Eq. [12] or [14] must,
in general, be solved iteratively until some error tolerance is met.
On the other hand, there has been interest over the years in linearly
implicit or non-iterative temporal approximations, which require
1262 Soil Science Society of America Journal
solving a fixed number of linear systems per time step (Paniconi et
al., 1991; Kavetski et al., 2002). Linearly implicit approximations
can be found in either multistep or multistage (e.g., Rosenbrock
schemes) contexts (Farthing et al., 2006). However they are still
much less common than fully implicit techniques.
Adaptive time discretizations for Richards’ equation are
much more common than adaptive spatial discretizations. A
classical approach is to increase or decrease the time step by a fixed
fraction based on the number of iterations taken by the nonlinear
solver (Paniconi and Putti, 1994). The methods introduced in
(Kavetski et al., 2001a, 2001b) and (Tocci et al., 1997) adapt the
time step size and/or order based on estimates of temporal error.
These approaches combine low and high-order approximations
to establish estimates of the local error associated with a solution
using a given time step and approximation order. They accept
or reject a solution based on user-selected tolerances and pick a
new order and/or time step size using heuristics often motivated
by simple control theory (Söderlind, 2006). These approaches
can be orders of magnitude more efficient than fixed time step
or heuristic alternatives, depending on the level of temporal
accuracy requested (Brenan et al., 1996; Kees and Miller, 2002;
Farthing et al., 2003b). On the other hand, heuristic time
stepping based on nonlinear solver performance may outperform
adaptation driven by local truncation error estimates for lower
accuracy regimes (D’Haese et al., 2007).
It is common to adjust time steps based on nonlinear solver
convergence as well as temporal truncation error estimates in
mature ODE and DAE solvers (Brenan et al., 1996; Hairer et al.,
1996). However, traditional approaches are not always well-suited
to highly nonlinear problems, especially when low temporal
accuracy is acceptable (Farthing et al., 2003b; Söderlind and
Wang, 2006). Efforts to combine nonlinear solver performance
and formal truncation error include Gustafsson and Söderlind
(1997), however, such approaches have not been evaluated
thoroughly for Richards’ equation and are not common practice.
Linear and Nonlinear Solvers
The desire to adjust a time step based on the performance of
the nonlinear solver arises from the fundamentally local nature of
Newton’s method and its variants that serve as the linearization
techniques for the vast majority of Richards’ equation solvers.
That is, returning to the notation for the discrete MOL system
at time level tn+1
G n+1 ( y n+1 ) = 0 [16]
Newton’s method attempts to solve Eq. [16] through a series of
linear problems
−G n+1 ( y ni +1 ) , y ni++11 =
J ni +1D y ni +1 = y ni +1 +D y ni +1 , i =
1,... [17]
∂G
until some convergence tolerance is met. Here J ni +1 =
∂ y
( y ni +1 )
is the Jacobian evaluated at y ni+1 Common convergence criteria
include specifying tolerances on the norm of the residual,
Gn+1(yn+1), or solution increment, Δyn+1. Classical convergence
results show that Newton’s method for Eq. [16] will converge
quadratically assuming that Gn+1 is sufficiently smooth and the
initial guess, y ni+1 is sufficiently close to the true solution (Kelley,
1995). As a result, a smaller time step, Δtn+1, often translates into
a better initial guess for Gn+1 and improved convergence. An
added benefit of smaller time steps is that the diagonal terms of
J are frequently proportional to 1⁄Δtn+1, and the linear system in
Eq. [17] becomes easier to solve as these diagonal terms become
more dominant (Söderlind, 2006).
Picard iteration and modified Picard iteration can be recast
as Newton’s method with an approximate Jacobian (Paniconi
and Putti, 1994) and so inherit this local nature. Like other
modified and inexact Newton’s methods (Kelley, 1995), they
trade reduced convergence rates in hope of gains in simplicity,
efficiency, and possibly robustness since the need for derivatives
of nonlinear soil constitutive relations is either eliminated
(classical Picard) or reduced (modified Picard) (Lehmann and
Ackerer, 1998). Another approach to reduce the impact of
highly nonlinear constitutive relations on Newton convergence
is to switch the primary variable between water content in drier
regions and capillary head elsewhere (Forsyth et al., 1995).
An interesting alternative to Newton or Picard methods is
the relaxation scheme used in Pop et al. (2004) as an approach
for linearizing mixed finite element approximations for Richards’
equation. It introduces a global, regularizing term into the
linearized problem for Gn+1 and exploits the monotonicity of soil-
moisture constitutive relationships to secure robust convergence
behavior with potentially less restrictive time step constraints
and better conditioned linear systems than Newton or modified
Picard approaches (Slodicka, 2002; List and Radu, 2016).
The issues associated with the local nature of Newton’s
method and its variants are not unique to Richards’ equation and
the need for additional modifications (so-called globalization
techniques) like line-search and trust-region methods (Kelley,
1995; Jones and Woodward, 2001; Knoll and Keyes, 2004;
Wang and Tchelepi, 2013) has been recognized in the broader
community. There has also been renewed interest in acceleration
techniques like Anderson acceleration (Walker and Ni, 2011),
which can be seen as an approach to improve the quality of
Picard (fixed-point) iterates by exploiting the nonlinear residual
history using Krylov subspace techniques (Lott et al., 2012).
Finally, a globalization approach that has been considered by
several researchers is to marry Newton’s method with a slower
converging technique like Picard iteration (Lehmann and
Ackerer, 1998; Bergamaschi et al., 1999) or the scheme from
Pop et al. (2004) to improve robustness for poor initial guesses
while recapturing quadratic convergence as intermediate iterates
approach the true solution (List and Radu, 2016).
As discussed above, the choice of time step naturally impacts
nonlinear solver performance (Gustafsson and Söderlind,
1997), and Richards’ equation solvers that use an implicit time
discretization must negotiate a balance between the size of time
step that can be taken and the difficulty of the resulting nonlinear
www.soils.org/publications/sssaj 1263
system. For large problems, the linear systems that must be solved
at each nonlinear iteration must be solved iteratively as well. The
most frequently used iterative linear solvers are Krylov based.
Briefly, Krylov subspace methods for the linear system Ax = b
search for a solution at iteration j in the Krylov space,
K j = span ( r0 , Ar0 , A 2 r0 , ..., A j−1r0 ) [18]
where r0 = b – Ax0. Conjugate Gradients are the canonical Krylov
method and can be related to a minimization problem over Κj
for a symmetric, positive definite A. For non-symmetric matrices,
which is typically the case with J unless Picard iteration is used,
methods like Generalized Minimal RESidual (GMRES) or Bi-
Conjugate Gradient STABilized (BiCGSTAB) are required (Saad
and van der Vorst, 2000). The time spent in solution of this system
is usually the dominant expense in the linearization process for
large problems (Tocci et al., 1998; Jones and Woodward, 2001)
A similar need to manage the interplay between the
temporal approximation, initial guess, and nonlinear problem,
G(n+1) = 0, plays out in the resulting linearized problem through
the conditioning of J. That is, direct (e.g., LU-based) solvers
are essentially global and seldom fail unless the linear problem
is extremely ill-conditioned. Sparse direct solvers saw much
improvement during the previous decade (Hénon et al., 2002; Li,
2005) and are practical for many two-dimensional problems and
moderate three-dimensional problems (Farthing et al., 2003a).
However, they are still unable to handle large, coupled three-
dimensional problems where iterative, Krylov-based techniques
are required (Knoll and Keyes, 2004). The performance of Krylov
methods in turn depends largely on effective preconditioning
techniques. There have been efforts to develop or apply robust,
scalable preconditioners for Richards’ equation (Jenkins et al., 2001;
Jones and Woodward, 2001), and high-quality software packages
like PETSc provide access to a number of modern preconditioners
as well as linear and nonlinear solvers (Balay et al., 1997). However,
uniformly robust preconditioning techniques and nonlinear solver
performance for Richards’ equation for remains an open research
topic (Juncu et al., 2011; Lipnikov et al., 2016).
For example, rather than attempt to extend or modify
existing preconditioners like geometric and algebraic multigrid
to the linearized systems arising from standard discretizations
of Richards’ equation, the work of Berninger et al. (2011) takes
an alternate view and modifies the underlying discretization to
realize the performance of multigrid for classical linear self-adjoint
problems. That is, introducing a Kirchoff (or matric flux potential)
transform for homogeneous media (Williams et al., 2000)
y
u ( y ) = ∫ K ( p ) dp [19]
−∞
has the effect of linearizing the gradient term in Darcy’s law
through the chain rule Ñu = K(y)Ñy. Together with temporally
explicit treatment of gravitational terms, this produces a system
that is equivalent to a convex minimization problem and for
which a class of nonlinear multigrid techniques are well-behaved
(Kornhuber, 2002). While this suffices for homogeneous
domains, additional conditions based on conserving mass
(via flux matching) and maintaining continuity of pressure at
material interfaces are required for heterogeneous domains. The
nonlinearity of these transmission conditions also requires a
more complex, nonlinear substructuring domain-decomposition
approach (Berninger et al., 2014).
Finally, for large problems with iterative solvers there is
the need to manage not just the impact of the time step on the
difficulty of the nonlinear system, but also its impact on the
resulting linear problems, which are often better conditioned for
smaller time steps. Moreover, clear efficiency gains are possible
by coordinating the linear and nonlinear solver tolerances
dynamically to avoid over-solving intermediate linear systems
(Tocci et al., 1998; Eisenstat and Walker, 1996).
Discussion and Alternatives
Computational inefficiency and lack of robustness in
Richards’ equation solvers used in practice most often manifest
as poor linear and nonlinear solver convergence and/or poor
performance of the time integrator. Our view is that a significant
portion of this poor convergence and instability arises from
under resolution in space of sharp fronts or other phenomena
where highly nonlinear soil properties change drastically over
neighboring elements or cells. However, the spatial component
of this behavior is not as readily apparent because estimates or
indicators of spatial accuracy are more difficult to obtain than
the convergence estimates and temporal error indicators that are
available for classical iterative linear and nonlinear techniques
and many time discretizations. Indeed, as discussed above, the
vast majority of spatial discretizations used in practice are based
on classical low-order finite element discretizations and cell-
centered finite differences and are combined with evaluation
techniques for soil constitutive relations like mass-lumping and
upwinding that are designed to favor robustness over accuracy.
One way to increase computational efficiency at the
expense of solution validity is to use resolutions considerably
coarser than the REV scale dictates. This requires the use of so-
called “effective” parameter values that diverge from physical
reality as the computational scale increases, and results in large
errors in the simulation’s representation of the physical system
even though it has converged. We hope instead that there will
be continued development and adoption of new spatial and
temporal discretization schemes. Following trends of maturing
computational fields, we expect to see improved, adaptive
algorithms driven by solution dynamics and a posteriori error
estimates (Mostaghimi et al., 2015; Baron et al., 2017). Moreover,
approaches are needed that consider entire solution error and
dynamics–for example, joint space-time discretization and error
control (Solin and Kuraz, 2011) rather than just a MOL approach
with temporal adaptivity but static spatial approximation
(Farthing et al., 2003b). While the implementation of full
spatially and temporally adaptive techniques is a clear hurdle,
1264 Soil Science Society of America Journal
the proper location of high-resolution space should also improve
robustness and accuracy in many cases.
Similarly, many of the issues with robustness of Richards’
equation solvers may be addressed with better understanding
of the dynamic interplay between temporal approximation and
linearization techniques. As discussed above, researchers have
tried to manage these dynamics through the time step (Tocci
et al., 1997; Kavetski et al., 2002; D’Haese et al., 2007) and/or
linearization technique (Paniconi and Putti, 1994; Lehmann and
Ackerer, 1998; Lipnikov et al., 2016). Alternative linearization
approaches (e.g., List and Radu, 2016) and better algorithms for
managing the complex dynamics of adaptive solvers taken, for
example, from control-theory (Bhaya and Kaszkurewicz, 2006)
could directly improve existing production codes and also lower
the barrier to adoption of more accurate spatial approximations.
We believe the notions of computational stability and
proportionality are useful concepts to guide the development
of modern Richards’ equation solvers. Simulators for Richards’
equation should have clearly defined input parameters for
controlling their accuracy, and changes in these parameters should
reliably increase (or decrease) accuracy and computational expense
at a predictable rate. In other words, tightening an error tolerance
slightly should lead to a commensurate decrease in error without
excessive increases in run time (Söderlind and Wang, 2006).
The idea of computational stability follows the line of
introducing mathematical concepts like continuous data
dependence and well-posedness to computational software. The
need for more rigorous computational science with standards
like reproducibility and repeatability of numerical experiments
has been recognized across a number of fields (Collberg and
Proebsting, 2016). It can be seen, for example, in efforts to
develop clear understanding of benchmarking, verification,
and validation for computational codes in fields like nuclear
engineering and aerospace engineering (Babuska and Oden,
2004; Oberkampf and Trucano, 2007). A key component of
such efforts is a suite of test problems identified and accepted
by the community that allows fair evaluation and comparison
of numerical techniques and simulation codes. While several
test problems recur throughout the literature and isolated
comparisons can be found, the community has lacked a central
repository of benchmark problems or recurring efforts to compare
results across research groups as can be found in other fields
(e.g., American Institute of Aeronautics and Astronautics, 1998;
Larson et al., 2014; American Society of Mechanical Engineers,
2017; International Association for the Engineering Modelling,
Analysis and Simulation Community, http://www.nafems.
org [verified 14 Sept. 2017]). The recent integrated hydrologic
model comparison project (Maxwell et al., 2014; Kollet et al.,
2017) and the International Soil Modeling Consortium (http://
soil-modeling.org; verified 10 Feb. 2017) represent welcome
steps in this direction. This is also an area where government
agencies (e.g., NOAA, the US National Weather Service, Office
of Water Prediction) could make a major contribution.
 Even as we work to improve the numerics, there continues
to be reason to evaluate Richards’ equation as a formulation
for unsaturated flow. Problems where gas-phase dynamics
are important usually demand a full two-phase formulation
(Helmig, 2011; Smits et al., 2013), while more recent work on
thermodynamically constrained averaging theory has =focused
on developing a consistent upscaling framework for deriving
extensions to Richards’ equation (Bronson, 2014; Gray and
Miller, 2014). These approaches are primarily aimed at extending
the operational range of validity for the mathematical system
as a representation for unsaturated flow (Or et al., 2015). In
some cases, new formulations or closure schemes may introduce
additional challenges for numerical schemes due to new terms
with additional, nonlinear coupling and complexity. On the
other hand, it is also possible that some numerical difficulties are can be solved using an explicit forward Euler scheme or other
unintended side effects of inadequate formulations rather than
being inherent to the unsaturated flow problem.
In a different direction, recent work to derive alternative
formulations that trade some of the generality of the traditional
Richards’ equation solution for gains in speed and reliability has
shown success. It is possible, for example, to convert Richards’
equation into a formulation for the evolution of soil moisture
front location with θ as the independent variable under suitable
assumptions (Philip, 1957; Swartzendruber, 1969). Ogden et al.
(2017) converted the one-dimensional Richards’ equation into a
form that they call the Soil Moisture Velocity Equation (SMVE)
that describes the speed of propagation of particular moisture
contents within a homogeneous soil layer:
 dz  −∂ K ( q )  ∂ y ( q )  ∂ 2 y/∂ z 2 [20]
=    −1  − D ( )
q recurring theme is the need to balance speed, robustness, and
 dt  q ∂q  ∂z  ∂ y/∂ z
Equation [20] is equivalent to the one-dimensional Richards
equation. This equation separates the flux due to gravity and the
integrated wetting front capillary drive from the flux caused
by the shape of the wetting front. The first term on the right-
hand side Eq. [20], which is called the “advection-like” term, was
derived by Ogden et al. (2015a). The second term on the right-
hand side of Eq. [20], called the “diffusion-like” term, is equal to numerical methods has largely been about reducing this tension
a diffusive flux due to the second spatial derivative of capillary
head divided by the slope of the capillary head along the wetting
front. This diffusion-like term will be zero when the gradient of
the capillarity is a constant along lines of constant θ, or in the
case of a sharp wetting front when ¶y/¶z = ¥. Both of these
cases represent situations that can cause convergence difficulties
in Richards’ equation solvers.
Fluxes calculated using the SMVE advection-like term agree
very closely with the numerical solution of Richards’ equation
(Ogden et al., 2015a), data from column experiments (Ogden et
al., 2015b) and exact analytical solutions (Ogden et al., 2017).
Results in all cases show that the diffusion-like term in Eq.
[20] is negligible unless one is interested in the exact shape of
the wetting front. Most importantly, the advection-like term of
the SMVE can be converted into an ODE using the MOL and
solved using a finite moisture-content discretization that divides
www.soils.org/publications/sssaj 1265
the vadose zone into N regions of thickness Δθ called “bins”
(Talbot and Ogden, 2008).
That resulting ODE for infiltrating water is (Ogden et
al., 2015a):
 dz  K ( q d ) − K ( q i )  y ( q d ) + hp 
   + 1  [21]
dt
 j q − q z 
d i  j 
where j is the bin index, θd is the moisture content of the bin
containing the highest moisture content, θi is the initial moisture
content, hp is the ponded depth if existent, and zj is the depth
to the wetting front in the jth finite moisture content bin. Note
that the ODE solution of Eq. [21] requires evaluation of no
spatial derivatives, eliminating what is a source of difficulty in
the traditional solution of Richards’ equation. Equation [21]
traditional ODE solutions such as Runge-Kutta. This ODE
is guaranteed to converge and to conserve mass using a finite
volume solution. While the SMVE solution method is at present
limited to one-dimensional infiltration in homogeneous soil
layers, the reliable, accurate and efficient ODE solution are major
advantages. Furthermore, Eq. [21] will solve sharp wetting fronts
with no difficulty and is not degenerate, making the method well
suited for use in Earth System and large-scale models of land-
atmosphere interaction.
Conclusions
We provided a high-level overview of numerical methods
for the solution of Richards’ equation with an eye toward
summarizing previous work and identifying major trends. A
accuracy. Historically, the dominant approaches have been low
order in space and time and used fixed spatial grids or meshes. This
has set up an unfortunate tension in many cases between accuracy
and robustness– or at least ease of convergence. Loosening
tolerances and coarsening resolution are expedient, but unreliable,
ways to improve speed or achieve convergence that have a negative
effect on solution accuracy and realism. The pursuit of improved
and increasing computational efficiency rigorously.
We hope to see continued development and research into
the numerical solution of Richards’ equation. Indeed, we believe
it is essential if application of Richards’ equation to large-scale
problems at valid discretizations is to become commonplace.
Successful Richards’ equation solvers in the future will likely
be adaptive in space and time, while efficiently managing the
interplay between discretizations on the one hand and linear
and/or nonlinear algebra solvers on the other. The recent
identification of the Soil Moisture Velocity Equation (Eq. [20])
indicates that, at least in the case of one-dimensional solutions
of Richards’ equation, there may be novel approaches yet to be
discovered. The most likely way for this development to continue
is to encourage the exchange of ideas and techniques with the
applied mathematics, computational mechanics, computer
science and Earth Science communities. Improved software

engineering and creation of archival verification and validation
data sets for use by the community are also essential if we are to
proceed rationally and make the most of our efforts.
Acknowledgments
Permission was granted by the Chief of Engineers to publish this
information. This material is based in part on work supported by
the National Science Foundation under Grant EAR-1360384 to the
University of Wyoming. Any opinions, findings, and conclusions or
recommendations expressed in this material are those of the author(s) and
do not necessarily reflect the views of the National Science Foundation.
References
Aavatsmark, I., T. Barkve, O. Bøe, and T. Mannseth. 1998. Discretization
on unstructured grids for inhomogeneous, anisotropic media. Part I:
Derivation of the methods. SIAM J. Sci. Comput. 19(5):1700–1716.
doi:10.1137/S1064827595293582
Allen, M.B., and C. Murphy. 1985. A finite element collocation method for
variably saturated flows in porous media. Numer. Methods Partial Differ.
Equ. 1:229–239. doi:10.1002/num.1690010306
Alt, W.H., and S. Luckhaus. 1983. Quasilinear elliptic-parabolic differential
equations. Math. Z. 183:311–341. doi:10.1007/BF01176474
American Institute of Aeronautics and Astronautics. 1998. AIAA guide for the
verification and validation of computational fluid dynamics simulations.
American Institute of Aeronautics & Astronautics, Reston, VA.
American Society of Mechanical Engineers. 2017. ASME V&V, Verification and
validation symposium. American Society of Mechanical Engineers, Las
Vegas, NV. 3–5 May 2017. https://www.asme.org/events/vandv (verified
6 Sept. 2017).
Arbogast, T., A. Todd, C.N. Dawson, P.T. Keenan, M.F. Wheeler, and Y. Ivan.
1998. Enhanced cell-centered finite differences for elliptic equations on
general geometry. SIAM J. Sci. Comput. 19(2):404–425. doi:10.1137/
S1064827594264545
Arbogast, T., A. Todd, and C. Zhangxin. 1995. On the implementation of mixed
methods as nonconforming methods for second-order elliptic problems.
Math. Comput. 64(211):943–972.
Arrarás, A., L. Portero, and J.C. Jorge. 2009. A combined mixed finite element
ADI scheme for solving Richards’ equation with mixed derivatives on
irregular grids. Appl. Numer. Math. 59(3-4):454–467. doi:10.1016/j.
apnum.2008.03.002
Assouline, S. 2013. Infiltration into soils: Conceptual approaches and solutions.
Water Resour. Res. 49:1755–1772. doi:10.1002/wrcr.20155
Babuska, I., and J.T. Oden. 2004. Verification and validation in computational
engineering and science: Basic concepts. Comput. Methods Appl. Mech.
Eng. 193(36-38):4057–4066. doi:10.1016/j.cma.2004.03.002
Balay, S., W.D. Gropp, L.C. McInnes, and B.F. Smith. 1997. Efficient management
of parallelism in object-oriented numerical software libraries. In: E. Arge,
A.M. Bruaset, and H.P. Langtengen, editors, Modern software tools for
scientific computing. Springer, New York. p. 163–202.
Baron, V., Y. Coudière, and P. Sochala. 2017. Adaptive multistep time
discretization and linearization based on a posteriori error estimates for
the Richards equation. Applied Numerical Mathematics 112:104–125.
doi:10.1016/j.apnum.2016.10.005
Barry, D.A., and G.C. Sander. 1991. Exact solutions for water infiltration with
an arbitrary surface flux or nonlinear solute adsorption. Water Resour. Res.
27:2667–2680. doi:10.1029/91WR01445
Barry, D.A., J.-Y. Parlange, G.C. Sander, and M. Sivapalan. 1993. A class of exact
solutions for Richards’ equation. J. Hydrol. 142:29–46. doi:10.1016/0022-
1694(93)90003-R
Bartlett, R.A. 2009. Integration strategies for computational science and
engineering software. In: 2009 ICSE Workshop on Software Engineering
for Computational Science and Engineering. Vancouver, BC, Canada. 23-
23 May 2009. doi:10.1109/SECSE.2009.5069160
Bause, M., and P. Knabner. 2004. Computation of variably saturated subsurface
flow by adaptive mixed hybrid finite element methods. Adv. Water Resour.
27(6):565–581. doi:10.1016/j.advwatres.2004.03.005
Belfort, B., B. Benjamin, R. Fanilo, Y. Anis, and L. François. 2009. An
efficient lumped mixed hybrid finite element formulation for variably
1266 Soil Science Society of America Journal
saturated groundwater flow. Vadose Zone J. 8(2):352–362. doi:10.2136/
vzj2008.0108
Bergamaschi, L., B. Luca, and M. Putti. 1999. Mixed finite elements and
Newton‐type linearizations for the solution of Richards equation. Int.
J. Numer. Methods Eng. 45(8):1025–1046. doi:10.1002/(SICI)1097-
0207(19990720)45:8<1025::AID-NME615>3.0.CO;2-G
Berninger, H., B. Heiko, K. Ralf, and S. Oliver. 2014. A multidomain
discretization of the Richards equation in layered soil. Comput. Geosci.
19(1):213–232. doi:10.1007/s10596-014-9461-8
Berninger, H., R. Kornhuber, and O. Sander. 2011. Fast and robust numerical
solution of the Richards equation in homogeneous soil. SIAM J. Numer.
Anal. 49(6):2576–2597. doi:10.1137/100782887
Bhaya, A., and E. Kaszkurewicz. 2006. Control perspectives on numerical
algorithms and matrix problems. Society for Industrial and Applied
Mathematics, Philadelphia, PA. doi:10.1137/1.9780898718669
Brenan, K.E., S.L. Campbell, and L.R. Petzold. 1996. Numerical solution
of initial-value problems in differential-algebraic equations.
Society for Industrial and Applied Mathematics, Philadelphia, PA.
doi:10.1137/1.9781611971224
Broadbridge, P., and I. White. 1988. Modelling solute transport, chemical
adsorption and cation exchange. In: Int. Hydrology and Water Resources
Symp. Nat. Conf. Publ. No. 92/19 (Preprints of Papers, Vol. 3:924-929).
2-4 October. Institution of Engineers, Perth, Australia.
Bronson, S. 2014. Assessing the minimum entropy production rate principle
for multiphase flow using the thermodynamically constrained averaging
theory approach. University of North Carolina, Chapel Hill.
Brunner, P., and C.T. Simmons. 2012. HydroGeoSphere: A fully integrated,
physically based hydrological model. Ground Water 50:170–176.
doi:10.1111/j.1745-6584.2011.00882.x
Campbell, J.C., J.M. Hyman, and M.J. Shashkov. 2002. Mimetic finite difference
operators for second-order tensors on unstructured grids. Comput. Math.
Appl. 44(1-2):157–173. doi:10.1016/S0898-1221(02)00137-2
Camporese, M., C. Paniconi, M. Putti and S. Orlandini. 2010. Surface-subsurface
flow modeling with path-based runoff routing, boundary condition-based
coupling, and assimilation of multisource observation data. Water Resour.
Res. 46. doi:10.1029/2008WR007536.
Celia, M.A., E.T. Bouloutas, and R.L. Zarba. 1990. A general mass-conservative
numerical solution for the unsaturated flow equation. Water Resour. Res.
26(7):1483–1496. doi:10.1029/WR026i007p01483
Chavent, G., G. Cohen, and J. Jaffre. 1984. Discontinuous upwinding and mixed
finite elements for two-phase flows in reservoir simulation. Computer
Methods in Applied Mechanics and Engineering 47:93–118.
Chavent, G., and J.E. Roberts. 1991. A unified physical presentation of mixed,
mixed-hybrid finite elements and standard finite difference approximations
for the determination of velocities in waterflow problems. Adv. Water
Resour. 14(6):329–348. doi:10.1016/0309-1708(91)90020-O
Chen, Z., G. Huan, and Y. Ma. 2006. Computational methods for multiphase
flows in porous media. Society for Industrial and Applied Mathematics,
Philadelphia, PA. doi:10.1137/1.9780898718942
Collberg, C., and T.A. Proebsting. 2016. Repeatability in computer systems
research. Commun. ACM 59(3):62–69.
Cox, P.M., R.A. Betts, C.B. Bunton, R.L.H. Essery, P.R. Rowntree, and J. Smith.
1999. The impact of new land surface physics on the GCM simulation
of climate and climate sensitivity. Clim. Dyn. 15:183–203. doi:10.1007/
s003820050276
Dawson, C., S. Sun, and M.F. Wheeler. 2004. Compatible algorithms for
coupled flow and transport. Comput. Methods Appl. Mech. Eng. 193(23-
26):2565–2580. doi:10.1016/j.cma.2003.12.059
D’Haese, C.M.F., M. Putti, C. Paniconi, and N.E.C. Verhoest. 2007. Assessment
of adaptive and heuristic time stepping for variably saturated flow. Int. J.
Numer. Methods Fluids 53(7):1173–1193. doi:10.1002/fld.1369
Diersch, H.J. 2013. Flow in variably saturated porous media. In: H.J. Diersch,
editor, FEFLOW: Finite Element Modeling of Flow, Mass and Heat
Transport in Porous and Fractured Media. Springer Verlag. Heidelberg,
Germany.
Downer, C.W., and F.L. Ogden. 2004. Appropriate vertical discretization of
Richards’ equation for two-dimensional watershed-scale modelling.
Hydrol. Processes 18:1–22. doi:10.1002/hyp.1306
Eisenstat, S.C., and H.F. Walker. 1996. Choosing the forcing terms in
an inexact Newton method. SIAM J. Sci. Comput. 17(1):16–32.
 doi:10.1137/0917003
Ek, M.B., K.E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, and J.D.
Tarpley. 2003. Implementation of Noah land surface model advances
in the National Centers for Environmental Prediction operational
mesoscale Eta model. J. Geophys. Res., D, Atmospheres 108(D22).
doi:10.1029/2002JD003296
Ern, A., and J.-L. Guermond. 2004. Theory and practice of finite elements.
Springer Verlag, Cambridge. doi:10.1007/978-1-4757-4355-5
Farthing, M.W., C.E. Kees, T.S. Coffey, C.T. Kelley, and C.T. Miller. 2003a.
Efficient steady-state solution techniques for variably saturated
groundwater flow. Adv. Water Resour. 26(8):833–849. doi:10.1016/
S0309-1708(03)00076-9
Farthing, M.W., C.E. Kees, and C.T. Miller. 2003b. Mixed finite element
methods and higher order temporal approximations for variably saturated
groundwater flow. Adv. Water Resour. 26(4):373–394. doi:10.1016/
S0309-1708(02)00187-2
Farthing, M.W., C.E. Kees, E.W. Jenkins, and C.T. Miller. 2006. An evaluation of
linearly implicit time discretization methods for approximating Richards’
equation. In: Binning, P., Engesgaard, P., Dahle, H., Pinder, G., Gray, W.G.,
editors, XVI International Conference on Computational Methods in
Water Resources (CMWR-XVI). Technical University of Denmark.
Fidkowski, K.J., and D.L. Darmofal. 2011. Review of output-based error
estimation and mesh adaptation in computational fluid dynamics. AIAA
J. 49(4):673–694. doi:10.2514/1.J050073
Finsterle, S., C. Doughty, M.B. Kowalsky, G.J. Moridis, L. Pan, T. Xu, et al. 2008.
Advanced vadose zone simulations using TOUGH. Vadose Zone J. 7:601–
609. doi:10.2136/vzj2007.0059
Forsyth, P.A., and M.C. Kropinski. 1997. Monotonicity considerations for
saturated–unsaturated subsurface flow. SIAM J. Sci. Comput. 18(5):1328–
1354. doi:10.1137/S1064827594265824
Forsyth, P.A., Y.S. Wu, and K. Pruess. 1995. Robust numerical methods for
saturated-unsaturated flow with dry initial conditions in heterogeneous
media. Adv. Water Resour. 18(1):25–38. doi:10.1016/0309-
1708(95)00020-J
Gerritsen, M.G. and L.J. Durlofsky. 2005. Modeling fluid flow in oil reservoirs.
Annual Review of Fluid Mechanics 37:211–238.
Gilding, B.H. 1991. Qualitative mathematical analysis of the Richards equation.
Transp. Porous Media 6(5-6):651–666. doi:10.1007/BF00137854
Gray, W.G., and S. Hassanizadeh. 1991. Paradoxes and realities in unsaturated flow
theory. Water Resour. Res. 27(8):1847–1854. doi:10.1029/91WR01259
Gray, W.G., and C.T. Miller. 2014. Introduction to the thermodynamically
constrained averaging theory for porous medium systems. Springer,
Cambridge. doi:10.1007/978-3-319-04010-3
Gustafsson, K., and G. Söderlind. 1997. Control strategies for the iterative
solution of nonlinear equations in ODE solvers. SIAM J. Sci. Comput.
18(1):23–40. doi:10.1137/S1064827595287109
Hairer, E., S.P. Nørsett, and G. Wanner. 1996. Solving ordinary differential
equations I: Nonstiff problems. Second ed. Springer Verlag, Cambridge.
doi:10.1007/978-3-642-05221-7
Healy, R.W. 2008. Simulating water, solute, and heat transport in the
subsurface with the VS2DI software package. Vadose Zone J. 7:632–639.
doi:10.2136/vzj2007.0075
Helmig, R. 2011. Multiphase flow and transport processes in the subsurface: A
contribution to the modeling of hydrosystems. Springer, Cambridge.
Hénon, P., P. Ramet, and J. Roman. 2002. PaStiX: A high-performance parallel
direct solver for sparse symmetric positive definite systems. Parallel
Comput. 28(2):301–321. doi:10.1016/S0167-8191(01)00141-7
Heroux, M.A., and J.M. Willenbring. 2009. Barely sufficient software
engineering: 10 practices to improve your CSE software. In: 2009 ICSE
Workshop on Software Engineering for Computational Science and
Engineering. doi:10.1109/SECSE.2009.5069157
Hoteit, H., and A. Firoozabadi. 2005. Multicomponent fluid flow by
discontinuous Galerkin and mixed methods in unfractured and fractured
media. Water Resour. Res. 41(11):W11412. doi:10.1029/2005WR004339
Howington, S.E., R.C. Berger, J.P. Hallberg, J.F. Peters, A.K. Stagg, E.W. Jenkins,
and C.T. Kelley. 1999. A model to simulate the interaction between
groundwater and surface water. ADA451802. US Army Engineering
Research and Development Center, Vicksburg, MS. p. 1-12.
Huber, R., and R. Helmig. 2000. Node-centered finite volume discretizations
for the numerical simulation of multiphase flow in heterogeneous porous
www.soils.org/publications/sssaj 1267
media. Comput. Geosci. 4:141–164. doi:10.1023/A:1011559916309
Jenkins, E.W., C.E. Kees, C.T. Kelley, and C.T. Miller. 2001. An aggregation-
based domain decomposition preconditioner for groundwater flow. SIAM
J. Sci. Comput. 23(2):430–441. doi:10.1137/S1064827500372274
Jones, J.E., and C.S. Woodward. 2001. Newton–Krylov-multigrid solvers for
large-scale, highly heterogeneous, variably saturated flow problems. Adv.
Water Resour. 24(7):763–774. doi:10.1016/S0309-1708(00)00075-0
Juncu, G., J. Gheorghe, N. Aurelian, and P. Constantin. 2011. Nonlinear
multigrid methods for numerical solution of the variably saturated flow
equation in two space dimensions. Transp. Porous Media 91(1):35–47.
doi:10.1007/s11242-011-9831-9
Kavetski, D., P. Binning, and S.W. Sloan. 2001a. Adaptive time stepping and
error control in a mass conservative numerical solution of the mixed form
of Richards equation. Adv. Water Resour. 24(6):595–605. doi:10.1016/
S0309-1708(00)00076-2
Kavetski, D., P. Binning, and S.W. Sloan. 2001b. Adaptive backward Euler
time stepping with truncation error control for numerical modelling of
unsaturated fluid flow. Int. J. Numer. Methods Eng. 53(6):1301–1322.
doi:10.1002/nme.329
Kavetski, D., P. Binning, and S.W. Sloan. 2002. Noniterative time stepping
schemes with adaptive truncation error control for the solution
of Richards equation. Water Resour. Res. 38(10):29-1–29-10.
doi:10.1029/2001WR000720
Kavetski, D., P. Binning, and S.W. Sloan. 2004. Truncation error and stability
analysis of iterative and non-iterative Thomas–Gladwell methods for
first-order non-linear differential equations. Int. J. Numer. Methods Eng.
60(12):2031–2043. doi:10.1002/nme.1035
Kees, C.E., M.W. Farthing, and C.N. Dawson. 2008. Locally conservative,
stabilized finite element methods for variably saturated flow. Comput.
Methods Appl. Mech. Eng. 197(51-52):4610–4625. doi:10.1016/j.
cma.2008.06.005
Kees, C.E., and C.T. Miller. 2002. Higher order time integration methods for
two-phase flow. Adv. Water Resour. 25(2):159–177. doi:10.1016/S0309-
1708(01)00054-9
Kelley, C.T. 1995. Iterative methods for linear and nonlinear equations. Society
for Industrial and Applied Mathematics, Philadelphia, PA.
Kirkland, M.R., R.G. Hills, and P.J. Wierenga. 1992. Algorithms for solving
Richards’ equation for variably saturated soils. Water Resour. Res.
28(8):2049–2058. doi:10.1029/92WR00802
Klausen, R.A., and T.F. Russell. 2004. Relationships among some locally
conservative discretization methods which handle discontinuous
coefficients. Comput. Geosci. 8(4):341–377. doi:10.1007/s10596-005-
1815-9
Klausen, R.A., F.A. Radu, and G.T. Eigestad. 2008. Convergence of MPFA on
triangulations and for Richards’ equation. Int. J. Numer. Methods Fluids
58(12):1327–1351. doi:10.1002/fld.1787
Knoll, D.A., and D.E. Keyes. 2004. Jacobian-free Newton–Krylov methods: A
survey of approaches and applications. J. Comput. Phys. 193(2):357–397.
doi:10.1016/j.jcp.2003.08.010
Kollet, S.J., and R.M. Maxwell. 2006. Integrated surface–groundwater
flow modeling: A free-surface overland flow boundary condition in a
parallel groundwater flow model. Adv. Water Resour. 29(7):945–958.
doi:10.1016/j.advwatres.2005.08.006
Kollet, S., M. Sulis, R.M. Maxwell, C. Paniconi, M. Putti, G. Bertoldi, E.T. Coon,
E. Cordano, S. Endrizzi, E. Kikinzon, E. Mouche, C. Mügler, Y.-J. Park,
J.C. Refsgaard, S. Stisen, and E. Sudicky. 2017. The integrated hydrologic
model intercomparison project, IH-MIP2: A second set of benchmark
results to diagnose integrated hydrology and feedbacks. Water Resour. Res.
doi:10.1002/2016WR019191
Kornhuber, R. 2002. On constrained Newton linearization and multigrid for
variational inequalities. Numer. Math. 91(4):699–721. doi:10.1007/
s002110100341
Kumar, M., C.J. Duffy, and K.M. Salvage. 2009. A second-order accurate, finite
volume–based, integrated hydrologic modeling (FIHM) framework
for simulation of surface and subsurface flow. Vadose Zone J. 8(4):873.
doi:10.2136/vzj2009.0014
Lang, J. 2013. Adaptive multilevel solution of nonlinear parabolic PDE systems:
Theory, algorithm, and applications. Springer Science & Business Media,
Berlin, Germany.
Larson, L., F. Stern, and M. Visonneau, editors. 2014. Numerical ship
 hydrodynamics: An assessment of the Gothenburg 2010 Workshop.
Springer Science & Business Media, Dordrecht, Germany.
Larson, M.G., and A.J. Niklasson. 2004. A conservative flux for the continuous
Galerkin method based on discontinuous enrichment. Calcolo 41(2):65–
76. doi:10.1007/s10092-004-0084-7
Lehmann, F., and P.H. Ackerer. 1998. Comparison of iterative methods for
improved solutions of the fluid flow equation in partially saturated porous
media. Transport in Porous Media 31:275–292.
Li, H., M.W. Farthing, C.N. Dawson, and C.T. Miller. 2007a. Local
discontinuous Galerkin approximations to Richards’ equation. Adv. Water
Resour. 30(3):555–575. doi:10.1016/j.advwatres.2006.04.011
Li, H., M.W. Farthing, and C.T. Miller. 2007b. Adaptive local discontinuous
Galerkin approximation to Richards’ equation. Adv. Water Resour.
30(9):1883–1901. doi:10.1016/j.advwatres.2007.02.007
Li, X.S. 2005. An overview of SuperLU. ACM Trans. Math. Softw. 31(3):302–
325. doi:10.1145/1089014.1089017
Liang, X., D.P. Lettenmaier, E.F. Wood, and S.J. Burges. 1994. A simple
hydrologically based model of land surface water and energy fluxes for
general circulation models. J. Geophys. Res. 99(D7):14,415–14,428.
doi:10.1029/94JD00483
Lipnikov, K., L. Konstantin, M. David, and S. Daniil. 2016. New preconditioning
strategy for Jacobian-free solvers for variably saturated flows with
Richards’ equation. Adv. Water Resour. 94:11–22. doi:10.1016/j.
advwatres.2016.04.016
List, F., and F.A. Radu. 2016. A study on iterative methods for solving Richards’
equation. Comput. Geosci. 20:341–353. doi:10.1007/s10596-016-9566-3
Lott, P.A., H.F. Walker, C.S. Woodward, and U.M. Yang. 2012. An accelerated
Picard method for nonlinear systems related to variably saturated flow.
Adv. Water Resour. 38:92–101. doi:10.1016/j.advwatres.2011.12.013
Manzini, G., and S. Ferraris. 2004. Mass-conservative finite volume methods
on 2-D unstructured grids for the Richards’ equation. Adv. Water Resour.
27(12):1199–1215. doi:10.1016/j.advwatres.2004.08.008
Maxwell, R.M., M. Putti, S. Meyerhoff, J.-O. Delfs, I.M. Ferguson, V. Ivanov,
J. Kim, O. Kolditz, S.J. Kollet, M. Kumar, S. Lopez, J. Niu, C. Paniconi,
Y.-J. Park, M.S. Phanikumar, C. Shen, E.A. Sudicky, and M. Sulis. 2014.
Surface-subsurface model intercomparison: A first set of benchmark
results to diagnose integrated hydrology and feedbacks. Water Resour. Res.
50(2):1531–1549. doi:10.1002/2013WR013725
McConnell, S. 2004. Code Complete: A practical handbook of software
construction. Microsoft Press, Redmond, WA.
Miller, C.T., C. Abhishek, and M.W. Farthing. 2006. A spatially and temporally
adaptive solution of Richards’ equation. Adv. Water Resour. 29(4):525–
545. doi:10.1016/j.advwatres.2005.06.008
Miller, C.T., C.N. Dawson, M.W. Farthing, T.Y. Hou, J.F. Huang, C.E. Kees,
et al. 2013. Numerical simulation of water resources problems: Models,
methods, and trends. Adv. Water Resour. 51:405–437. doi:10.1016/j.
advwatres.2012.05.008
Miller, C.T., C. George, P.T. Imhoff, J.F. McBride, J.A. Pedit, and J.A. Trangenstein.
1998. Multiphase flow and transport modeling in heterogeneous porous
media: Challenges and approaches. Adv. Water Resour. 21(2):77–120.
doi:10.1016/S0309-1708(96)00036-X
Mostaghimi, P., J.R. Percival, D. Pavlidis, R.J. Ferrier, J.L.M.A. Gomes, G.J.
Gorman, et al. 2015. Anisotropic Mesh Adaptivity and Control Volume
Finite Element Methods for Numerical Simulation of Multiphase Flow
in Porous Media. Math. Geosci. 47:417–440. doi:10.1007/s11004-014-
9579-1
Niswonger, R.G., and D.E. Prudic. 2009. Comment on “Evaluating interactions
between groundwater and vadose zone using the HYDRUS-based flow
package for MODFLOW”. Vadose Zone J. 8:818–819. doi:10.2136/
vzj2008.0155
Niu, G.-Y., Z.L. Yang, R.E. Dickinson, L.E. Gulden, and H. Su. 2007.
Development of a simple groundwater model for use in climate models
and evaluation with Gravity Recovery and Climate Experiment data. J.
Geophys. Res. 112:D07103. doi:10.1029/2006JD007522
Oberkampf, W.L., and T.G. Trucano. 2007. Verification and validation
benchmarks. Sandia National Laboratory, Albuquerque, NM.
doi:10.2172/901974
Ogden, F.L., W. Lai, R.C. Steinke, J. Zhu, C.A. Talbot, and J.L. Wilson. 2015a.
A new general 1-D vadose zone flow solution method. Water Resour. Res.
51:4282–4300. doi:10.1002/2015WR017126
1268 Soil Science Society of America Journal
Ogden, F.L., W. Lai, R.C. Steinke, and J. Zhu. 2015b. Validation of finite water-
content vadose zone dynamics method using column experiments with
a moving water table and applied surface flux. Water Resour. Res. 51.
doi:10.1002/2014WR016454
Ogden, F.L., M.B. Allen, W. Lai, J. Zhu, M. Seo, C.C. Douglas, and C.A. Talbot.
2017. The soil moisture velocity equation. J. Adv. Model. Earth Syst.
doi:10.1002/2017MS000931
Or, D., P. Lehmann, and S. Assouline. 2015. Natural length scales define the
range of applicability of the Richards equation for capillary flows. Water
Resour. Res. 51:7130–7144. doi:10.1002/2015WR017034
Orgogozo, L., N. Renon, C. Soulaine, F. Henon, S.K. Tomer, D. Labat, et al.
2014. An open source massively parallel solver for Richards equation:
Mechanistic modelling of water fluxes at the watershed scale. Comput.
Phys. Commun. 185:3358–3371. doi:10.1016/j.cpc.2014.08.004
Pain, C.C., A.P. Umpleby, C.R.E. de Oliveira, and A.J.H. Goddard. 2001.
Tetrahedral mesh optimisation and adaptivity for steady-state and
transient finite element calculations. Comput. Methods Appl. Mech. Eng.
190:3771–3796. doi:10.1016/S0045-7825(00)00294-2
Paniconi, C., P. Claudio, A.A. Aldama, and E.F. Wood. 1991. Numerical
evaluation of iterative and noniterative methods for the solution of the
nonlinear Richards equation. Water Resour. Res. 27(6):1147–1163.
doi:10.1029/91WR00334
Paniconi, C., and M. Putti. 1994. A comparison of Picard and Newton
iteration in the numerical solution of multidimensional variably
saturated flow problems. Water Resour. Res. 30(12):3357–3374.
doi:10.1029/94WR02046
Paniconi, C., and M. Putti. 2015. Physically based modeling in catchment
hydrology at 50: Survey and outlook. Water Resour. Res. 51:7090–7129.
doi:10.1002/2015WR017780
Pettway, J.S., J.H. Schmidt, and A.K. Stagg. 2010. Adaptive meshing in a mixed
regime hydrologic simulation model. Comput. Geosci. 14(4):665–674.
doi:10.1007/s10596-010-9179-1
Philip, J.R. 1957. Numerical solution of equations of the diffusion type
with diffusivity concentration-dependent. II. Aust. J. Phys. 10:29–42.
doi:10.1071/PH570029
Pop, I.S., F. Radu, and P. Knabner. 2004. Mixed finite elements for the Richards’
equation: Linearization procedure. J. Comput. Appl. Math. 168(1-
2):365–373. doi:10.1016/j.cam.2003.04.008
Rannacher, R. 2001. Adaptive Galerkin finite element methods for partial
differential equations. J. Comput. Appl. Math. 128(1-2):205–233.
doi:10.1016/S0377-0427(00)00513-6
Rathfelder, K., and L.M. Abriola. 1994. Mass conservative numerical solutions
of the head-based Richards equation. Water Resour. Res. 30:2579–2586.
Richards, L.A. 1931. Capillary conduction of liquids through porous medium.
Physics 1:318–333.
Richardson, L.F. 1922. Weather prediction by numerical process. University
Press, Cambridge. p. 262.
Rogers, C., M.P. Stallybrass, and D.L. Clements. 1983. On two phase filtration
under gravity and with boundary infiltration: Application of a Backlund
transformation. Nonlin. Anal. Theory Meth. Appl. 7(7):785–799.
doi:10.1016/0362-546X(83)90034-2
Ross, P.J., and J.-Y. Parlange. 1994. Comparing exact and numerical solutions of
Richards’ equation for one-dimensional infiltration and drainage. Soil Sci.
157(6):341–344. doi:10.1097/00010694-199406000-00002
Saad, Y., and H.A. van der Vorst. 2000. Iterative solution of linear systems in
the 20th century. J. Comput. Appl. Math. 123:1–33. doi:10.1016/S0377-
0427(00)00412-X
Sander, G.C., J.-Y. Parlange, V. Kuhnel, W.L. Hogarth, D. Lockington, and J.P.J.
O’Kane. 1988. Exact nonlinear solution for constant flux infiltration. J.
Hydrol. 97:341–346. doi:10.1016/0022-1694(88)90123-0
Schwab, C. 1998. P- and Hp- finite element methods: Theory and applications in
solid and fluid mechanics. Univ. Press, Oxford.
Scudeler, C., M. Putti, and C. Paniconi. 2016. Mass-conservative reconstruction
of Galerkin velocity fields for transport simulations. Adv. Water Resour.
94:470–485. doi:10.1016/j.advwatres.2016.06.011
Selker, J., and S. John. 2004. Modelling variably saturated flow with HYDRUS-
2D. Vadose Zone J. 3(2):725. doi:10.2136/vzj2004.0725
Shewchuk, J.R. 2002. What is a good linear finite element?- Interpolation,
Conditioning, Anisotropy, and Quality Measures. Proceedings of the 11th
International Meshing Roundtable. Sandia National Laboratories and
 Cornell University, Ithaca, New York.
Simunek, J., M.Th. van Genuchten, and M. Sejna. 2005. HYDRUS-1D software
package for simulating the movement of water, heat, and multiple solutes
in variably saturated media, version 3.0, HYDRUS software series
1. Department of Environmental Sciences, University of California
Riverside, Riverside, CA.
Simunek, J., M.T. van Genuchten, and M. Sejna. 2008. Development and
applications of the HYDRUS and STANMOD software packages and
related codes. Vadose Zone J. 7:587–600. doi:10.2136/vzj2007.0077
Slodicka, M. 2002. A robust and efficient linearization scheme for doubly
nonlinear and degenerate parabolic problems arising in flow in porous
media. SIAM J. Sci. Comput. 23(5):1593–1614. doi:10.1137/
S1064827500381860
Smits, K.M., A. Cihan, T. Sakaki, S.E. Howington, J.F. Peters, and T.H.
Illangasekare. 2013. Soil moisture and thermal behavior in the vicinity
of buried objects affecting remote sensing detection: Experimental and
modeling investigation. IEEE Trans. Geosci. Rem. Sens. 51(5):2675–
2688. doi:10.1109/TGRS.2012.2214485
Söderlind, G. 2006. Time-step selection algorithms: Adaptivity, control, and
signal processing. Appl. Numer. Math. 56(3-4):488–502. doi:10.1016/j.
apnum.2005.04.026
Söderlind, G., and L. Wang. 2006. Adaptive time-stepping and computational
stability. J. Comput. Appl. Math. 185(2):225–243. doi:10.1016/j.
cam.2005.03.008
Solin, P., and M. Kuraz. 2011. Solving the nonstationary Richards equation with
adaptive hp-FEM. Adv. Water Resour. 34(9):1062–1081. doi:10.1016/j.
advwatres.2011.04.020
Sun, S., and M.F. Wheeler. 2006. Projections of velocity data for the compatibility
with transport. Comput. Methods Appl. Mech. Eng. 195(7-8):653–673.
doi:10.1016/j.cma.2005.02.011
Swartzendruber, D. 1969. The flow of water in unsaturated soils. In: R.J.M. de
Wiest, editor, Flow through porous media. Academic Press, New York. p.
215–292.
Talbot, C.A., and F.L. Ogden. 2008. A method for computing infiltration and
redistribution in a discretized moisture content domain. Water Resour.
Res. 44(8):W08453. doi:10.1029/2008WR006815
Tocci, M.D., C.T. Kelley, and C.T. Miller. 1997. Accurate and economical solution
of the pressure-head form of Richards’ equation by the method of lines.
Adv. Water Resour. 20(1):1–14. doi:10.1016/S0309-1708(96)00008-5
Tocci, M.D., C.T. Kelley, C.T. Miller, and C.E. Kees. 1998. Inexact
Newton methods and the method of lines for solving Richards’
equation in two space dimensions. Comput. Geosci. 2(4):291–309.
doi:10.1023/A:1011562522244
Trangenstein, J.A. 2013. Numerical solution of elliptic and parabolic partial
differential equations with CD-ROM. Univ. Press, Cambridge.
doi:10.1017/CBO9781139025508
Trefry, M.G., and C. Muffels. 2007. Feflow: A finite-element ground water flow
and transport modeling tool. Ground Water 45:525–528. doi:10.1111/
j.1745-6584.2007.00358.x
www.soils.org/publications/sssaj 1269
van Dam, J.C., and R.A. Feddes. 2000. Numerical simulation of infiltration,
evaporation, and shallow groundwater levels with the Richards’ equation.
J. Hydrol. 233(1-4):72–85. doi:10.1016/S0022-1694(00)00227-4
Van Duyn, C.J., and L.A. Peletier. 1982. Nonstationary filtration in partially
saturated porous media. Arch. Ration. Mech. Anal. 78(2):173–198.
doi:10.1007/BF00250838
van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44(5):892.
doi:10.2136/sssaj1980.03615995004400050002x
Vereecken, H., A. Schnepf, J.W. Hopmans, M. Javaux, D. Or, T. Roose, et al. 2016.
Modeling soil processes: Review, key challenges, and new perspectives.
Vadose Zone J. 15. doi:10.2136/vzj2015.09.0131
Vogel, H.-J., and O. Ippisch. 2008. Estimation of a critical spatial discretization
limit for solving Richards’ equation at large scales. Vadose Zone J.
7(1):112–114. doi:10.2136/vzj2006.0182
Vogel, T., M.T. van Genuchten, and M. Cislerova. 2001. Effect of the shape of
the soil hydraulic functions near saturation on variably-saturated flow
predictions. Adv. Water Resour. 24:133–144.
Walker, H.F., and P. Ni. 2011. Anderson acceleration for fixed-point iterations.
SIAM J. Numer. Anal. 49(4):1715–1735. doi:10.1137/10078356X
Wang, X., and H.A. Tchelepi. 2013. Trust-region based solver for nonlinear
transport in heterogeneous porous media. J. Comput. Phys. 253:114–137.
doi:10.1016/j.jcp.2013.06.041
Williams, G.A., C.T. Miller, and C.T. Kelley. 2000. Transformation approaches
for simulating flow in variably saturated porous media. Water Resour. Res.
36(4):923–934. doi:10.1029/1999WR900349
Woodward, C.S., and C.N. Dawson. 2000. Analysis of expanded mixed finite
element methods for a nonlinear parabolic equation modeling flow into
variably saturated porous media. SIAM J. Numer. Anal. 37(3):701–724.
doi:10.1137/S0036142996311040
Yang, Z.L., G.Y. Niu, K.E. Mitchell, F. Chen, M.B. Ek, M. Barlage, and Y. Xia.
2011. The community Noah land surface model with multiparameterization
options (Noah-MP): 2. Evaluation over global river basins. J. Geophys. Res.
Atmos. 116(D12). doi:10.1029/2010JD015140
Yeh, G.-T., D.-S. Shih, and J.-R.C. Cheng. 2011. An integrated media, integrated
processes watershed model. Comput. Fluids 45(1):2–13. doi:10.1016/j.
compfluid.2010.11.018
Younes, A., M. Fahs, and B. Belfort. 2013. Monotonicity of the cell-centred
triangular MPFA method for saturated and unsaturated flow in
heterogeneous porous media. J. Hydrol. 504:132–141. doi:10.1016/j.
jhydrol.2013.09.041
Zha, Y., J. Yang, L. Yin, Y. Zhang, W. Zeng, and L. Shi. 2017. A modified
Picard iteration scheme for overcoming numerical difficulties of
simulating infiltration into dry soil. J. Hydrol. 551:56–69. doi:10.1016/j.
jhydrol.2017.05.053
Zyvoloski, G. 2007. FEHM: A control volume finite element code for simulating
subsurface multi-phase multi-fluid heat and mass transfer. LAUR-07-
3359. Los Alamos National Laboratory, Los Alamos, NM.